Recent evidence has shown that non-coding RNAs are ubiquitous in the cell and that their functions and structure vary to a greater extent than previously imagined. Multiple new RNA classes have been implicated in many diseases, and understanding how these RNAs work is a critical need. While exciting discoveries are accumulating, our functional knowledge of these new RNAs remains limited. Here we propose to couple a new high-throughput RNA duplex sequencing technology with new, computational methods to economically study novel functional non-coding RNA at a genomic scale. We propose to develop two computational methodologies to characterize putative newly found non-coding RNAs on the genomic scale. First, we will develop a maximum likelihood approach that estimates RNA secondary structure using RNA-seq assays that preferentially sequence single- or double-stranded nucleotides. Second, we will develop a machine-learning framework that predicts the functional category of novel non-coding RNAs using length and structure features of known RNAs. These structural and functional predictions will be validated by comparative genomics and experimentation. We will develop databases and analysis software, and investigate the human genome and five other model organisms. In total, our findings will yield tremendous insights into non-coding RNA biology and will substantially impact continued study of these important molecules.